光谱技术在谷物水分检测中的应用研究

Research on the Application of Spectral Technology in Grain Moisture Detection

摘要: 谷物是全球口粮的主体，其水分含量又是决定收获时机、储藏稳定性、加工品质与食用安全的关键指标。传统测定方法虽准确度可靠，却难以满足收购、仓储与加工环节的快速在线检测需求；相比之下，光谱检测无需试剂与前处理，可在整粒谷物上直接测定，近红外与中红外还能同时给出蛋白、脂肪等指标。本文从原理、检测方式与优缺点三方面比较近红外光谱、中红外光谱、高光谱成像、太赫兹与微波介电五类技术在小麦、玉米、水稻、大豆等谷物水分检测中的应用，梳理光谱预处理、特征波长选择、定量建模等化学计量学方法，并说明可解释人工智能用于识别关键波段的做法。

Abstract: Cereals form the staple of the global food supply, and their moisture content is a key indicator governing harvest timing, storage stability, processing quality and food safety. Conventional determination methods give adequate accuracy, yet cannot meet the demand for rapid on-line determination at purchasing, storage and processing stages; in contrast, spectroscopic methods need no reagents or sample pretreatment and can measure intact kernels directly, and near-infrared and mid-infrared spectroscopy can additionally report protein and fat contents. This review compares five techniques (near-infrared spectroscopy, mid-infrared spectroscopy, hyperspectral imaging, terahertz, and microwave dielectric) for moisture detection in wheat, corn, rice, soybean and other grains, in terms of detection principles, detection modes, and advantages and disadvantages. It also outlines chemometric methods including spectral preprocessing, characteristic wavelength selection and quantitative modeling, and describes how explainable artificial intelligence is used to identify key spectral bands.